radeon/r200/r300: cleanup some of the renderbuffer code
[mesa.git] / src / mesa / drivers / dri / i965 / brw_wm.c
1 /*
2 Copyright (C) Intel Corp. 2006. All Rights Reserved.
3 Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
4 develop this 3D driver.
5
6 Permission is hereby granted, free of charge, to any person obtaining
7 a copy of this software and associated documentation files (the
8 "Software"), to deal in the Software without restriction, including
9 without limitation the rights to use, copy, modify, merge, publish,
10 distribute, sublicense, and/or sell copies of the Software, and to
11 permit persons to whom the Software is furnished to do so, subject to
12 the following conditions:
13
14 The above copyright notice and this permission notice (including the
15 next paragraph) shall be included in all copies or substantial
16 portions of the Software.
17
18 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
19 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
21 IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
22 LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
23 OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
24 WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
25
26 **********************************************************************/
27 /*
28 * Authors:
29 * Keith Whitwell <keith@tungstengraphics.com>
30 */
31
32 #include "main/texformat.h"
33 #include "brw_context.h"
34 #include "brw_util.h"
35 #include "brw_wm.h"
36 #include "brw_state.h"
37
38
39 /** Return number of src args for given instruction */
40 GLuint brw_wm_nr_args( GLuint opcode )
41 {
42 switch (opcode) {
43 case WM_PIXELXY:
44 case WM_CINTERP:
45 case WM_WPOSXY:
46 return 1;
47 case WM_LINTERP:
48 case WM_DELTAXY:
49 case WM_PIXELW:
50 return 2;
51 case WM_FB_WRITE:
52 case WM_PINTERP:
53 return 3;
54 default:
55 assert(opcode < MAX_OPCODE);
56 return _mesa_num_inst_src_regs(opcode);
57 }
58 }
59
60
61 GLuint brw_wm_is_scalar_result( GLuint opcode )
62 {
63 switch (opcode) {
64 case OPCODE_COS:
65 case OPCODE_EX2:
66 case OPCODE_LG2:
67 case OPCODE_POW:
68 case OPCODE_RCP:
69 case OPCODE_RSQ:
70 case OPCODE_SIN:
71 case OPCODE_DP3:
72 case OPCODE_DP4:
73 case OPCODE_DPH:
74 case OPCODE_DST:
75 return 1;
76
77 default:
78 return 0;
79 }
80 }
81
82
83 static void do_wm_prog( struct brw_context *brw,
84 struct brw_fragment_program *fp,
85 struct brw_wm_prog_key *key)
86 {
87 struct brw_wm_compile *c;
88 const GLuint *program;
89 GLuint program_size;
90
91 c = brw->wm.compile_data;
92 if (c == NULL) {
93 brw->wm.compile_data = calloc(1, sizeof(*brw->wm.compile_data));
94 c = brw->wm.compile_data;
95 } else {
96 memset(c, 0, sizeof(*brw->wm.compile_data));
97 }
98 memcpy(&c->key, key, sizeof(*key));
99
100 c->fp = fp;
101 c->env_param = brw->intel.ctx.FragmentProgram.Parameters;
102
103 brw_init_compile(brw, &c->func);
104 if (brw_wm_is_glsl(&c->fp->program)) {
105 brw_wm_glsl_emit(brw, c);
106 } else {
107 /* Augment fragment program. Add instructions for pre- and
108 * post-fragment-program tasks such as interpolation and fogging.
109 */
110 brw_wm_pass_fp(c);
111
112 /* Translate to intermediate representation. Build register usage
113 * chains.
114 */
115 brw_wm_pass0(c);
116
117 /* Dead code removal.
118 */
119 brw_wm_pass1(c);
120
121 /* Register allocation.
122 */
123 c->grf_limit = BRW_WM_MAX_GRF/2;
124
125 brw_wm_pass2(c);
126
127 c->prog_data.total_grf = c->max_wm_grf;
128 if (c->last_scratch) {
129 c->prog_data.total_scratch =
130 c->last_scratch + 0x40;
131 } else {
132 c->prog_data.total_scratch = 0;
133 }
134
135 /* Emit GEN4 code.
136 */
137 brw_wm_emit(c);
138 }
139 if (INTEL_DEBUG & DEBUG_WM)
140 fprintf(stderr, "\n");
141
142 /* get the program
143 */
144 program = brw_get_program(&c->func, &program_size);
145
146 dri_bo_unreference(brw->wm.prog_bo);
147 brw->wm.prog_bo = brw_upload_cache( &brw->cache, BRW_WM_PROG,
148 &c->key, sizeof(c->key),
149 NULL, 0,
150 program, program_size,
151 &c->prog_data,
152 &brw->wm.prog_data );
153 }
154
155
156
157 static void brw_wm_populate_key( struct brw_context *brw,
158 struct brw_wm_prog_key *key )
159 {
160 GLcontext *ctx = &brw->intel.ctx;
161 /* BRW_NEW_FRAGMENT_PROGRAM */
162 struct brw_fragment_program *fp =
163 (struct brw_fragment_program *)brw->fragment_program;
164 GLuint lookup = 0;
165 GLuint line_aa;
166 GLuint i;
167
168 memset(key, 0, sizeof(*key));
169
170 /* Build the index for table lookup
171 */
172 /* _NEW_COLOR */
173 if (fp->program.UsesKill ||
174 ctx->Color.AlphaEnabled)
175 lookup |= IZ_PS_KILL_ALPHATEST_BIT;
176
177 if (fp->program.Base.OutputsWritten & (1<<FRAG_RESULT_DEPR))
178 lookup |= IZ_PS_COMPUTES_DEPTH_BIT;
179
180 /* _NEW_DEPTH */
181 if (ctx->Depth.Test)
182 lookup |= IZ_DEPTH_TEST_ENABLE_BIT;
183
184 if (ctx->Depth.Test &&
185 ctx->Depth.Mask) /* ?? */
186 lookup |= IZ_DEPTH_WRITE_ENABLE_BIT;
187
188 /* _NEW_STENCIL */
189 if (ctx->Stencil.Enabled) {
190 lookup |= IZ_STENCIL_TEST_ENABLE_BIT;
191
192 if (ctx->Stencil.WriteMask[0] ||
193 ctx->Stencil.WriteMask[ctx->Stencil._BackFace])
194 lookup |= IZ_STENCIL_WRITE_ENABLE_BIT;
195 }
196
197 line_aa = AA_NEVER;
198
199 /* _NEW_LINE, _NEW_POLYGON, BRW_NEW_REDUCED_PRIMITIVE */
200 if (ctx->Line.SmoothFlag) {
201 if (brw->intel.reduced_primitive == GL_LINES) {
202 line_aa = AA_ALWAYS;
203 }
204 else if (brw->intel.reduced_primitive == GL_TRIANGLES) {
205 if (ctx->Polygon.FrontMode == GL_LINE) {
206 line_aa = AA_SOMETIMES;
207
208 if (ctx->Polygon.BackMode == GL_LINE ||
209 (ctx->Polygon.CullFlag &&
210 ctx->Polygon.CullFaceMode == GL_BACK))
211 line_aa = AA_ALWAYS;
212 }
213 else if (ctx->Polygon.BackMode == GL_LINE) {
214 line_aa = AA_SOMETIMES;
215
216 if ((ctx->Polygon.CullFlag &&
217 ctx->Polygon.CullFaceMode == GL_FRONT))
218 line_aa = AA_ALWAYS;
219 }
220 }
221 }
222
223 brw_wm_lookup_iz(line_aa,
224 lookup,
225 key);
226
227
228 /* BRW_NEW_WM_INPUT_DIMENSIONS */
229 key->projtex_mask = brw->wm.input_size_masks[4-1] >> (FRAG_ATTRIB_TEX0 - FRAG_ATTRIB_WPOS);
230
231 /* _NEW_LIGHT */
232 key->flat_shade = (ctx->Light.ShadeModel == GL_FLAT);
233
234 /* _NEW_TEXTURE */
235 for (i = 0; i < BRW_MAX_TEX_UNIT; i++) {
236 const struct gl_texture_unit *unit = &ctx->Texture.Unit[i];
237
238 if (unit->_ReallyEnabled) {
239 const struct gl_texture_object *t = unit->_Current;
240 const struct gl_texture_image *img = t->Image[0][t->BaseLevel];
241 if (img->InternalFormat == GL_YCBCR_MESA) {
242 key->yuvtex_mask |= 1 << i;
243 if (img->TexFormat->MesaFormat == MESA_FORMAT_YCBCR)
244 key->yuvtex_swap_mask |= 1 << i;
245 }
246
247 key->tex_swizzles[i] = t->_Swizzle;
248 }
249 else {
250 key->tex_swizzles[i] = SWIZZLE_NOOP;
251 }
252 }
253
254 /* Shadow */
255 key->shadowtex_mask = fp->program.Base.ShadowSamplers;
256
257 /* _NEW_BUFFERS */
258 /*
259 * Include the draw buffer origin and height so that we can calculate
260 * fragment position values relative to the bottom left of the drawable,
261 * from the incoming screen origin relative position we get as part of our
262 * payload.
263 *
264 * We could avoid recompiling by including this as a constant referenced by
265 * our program, but if we were to do that it would also be nice to handle
266 * getting that constant updated at batchbuffer submit time (when we
267 * hold the lock and know where the buffer really is) rather than at emit
268 * time when we don't hold the lock and are just guessing. We could also
269 * just avoid using this as key data if the program doesn't use
270 * fragment.position.
271 *
272 * This pretty much becomes moot with DRI2 and redirected buffers anyway,
273 * as our origins will always be zero then.
274 */
275 if (brw->intel.driDrawable != NULL) {
276 key->origin_x = brw->intel.driDrawable->x;
277 key->origin_y = brw->intel.driDrawable->y;
278 key->drawable_height = brw->intel.driDrawable->h;
279 }
280
281 /* Extra info:
282 */
283 key->program_string_id = fp->id;
284
285 }
286
287
288 static void brw_prepare_wm_prog(struct brw_context *brw)
289 {
290 struct brw_wm_prog_key key;
291 struct brw_fragment_program *fp = (struct brw_fragment_program *)
292 brw->fragment_program;
293
294 brw_wm_populate_key(brw, &key);
295
296 /* Make an early check for the key.
297 */
298 dri_bo_unreference(brw->wm.prog_bo);
299 brw->wm.prog_bo = brw_search_cache(&brw->cache, BRW_WM_PROG,
300 &key, sizeof(key),
301 NULL, 0,
302 &brw->wm.prog_data);
303 if (brw->wm.prog_bo == NULL)
304 do_wm_prog(brw, fp, &key);
305 }
306
307
308 /* See brw_wm.c:
309 */
310 const struct brw_tracked_state brw_wm_prog = {
311 .dirty = {
312 .mesa = (_NEW_COLOR |
313 _NEW_DEPTH |
314 _NEW_STENCIL |
315 _NEW_POLYGON |
316 _NEW_LINE |
317 _NEW_LIGHT |
318 _NEW_BUFFERS |
319 _NEW_TEXTURE),
320 .brw = (BRW_NEW_FRAGMENT_PROGRAM |
321 BRW_NEW_WM_INPUT_DIMENSIONS |
322 BRW_NEW_REDUCED_PRIMITIVE),
323 .cache = 0
324 },
325 .prepare = brw_prepare_wm_prog
326 };
327